Kai-Fu Lee publicly abandons pretraining of ultra-large models
Publicly declared an end to pretraining ultra-large-parameter models — said only big companies can afford to burn through giant models, and that a startup's way to survive is not burning more GPUs but making every yuan count. Staff for ultra-large-model training were moved to a joint lab with Alibaba Cloud
Why it matters
The first public declaration of a change of course among China's "six little tigers". The point at which startups giving up frontier pretraining became an established narrative
Source — Primary sources first; secondary reports are labelled “Press”
Pressnbd.com.cnhttps://www.nbd.com.cn/articles/2025-01-07/3713092.htmlAdded in Week 35, 2026 · See that week